About the Role
You'll own the connection between human and robot: from the GELLO leader device to the KUKA arm. Real-time control, servo communication, inverse kinematics, and making teleoperation feel instant. This requires someone who understands the hardware-to-software boundary—motor control, industrial protocols, and real-time systems.
What You'll Do
- Build the GELLO-to-KUKA teleoperation stack: read leader device at 50Hz, interpolate to 250Hz, send to KUKA RSI
- Implement communication with serial bus servos (Zhongling, Feetech) for custom leader devices
- Design interpolation and filtering algorithms for smooth, responsive robot motion
- Integrate force/torque sensing for haptic feedback
- Debug latency issues across the full stack: USB latency, kernel scheduling, network jitter, RSI timing
- Work with the AI engineer to ensure teleoperation data is captured correctly for training
What We're Looking For
- Proficiency in low-level systems languages: C, C++ (required), Rust or Go (bonus)
- Deep understanding of low-level systems concepts: multithreading, memory management, real-time scheduling
- Design, debug, and integrate low-level communication protocols: RS-485, CAN, UART, SPI
- Experience with real-time embedded systems (FreeRTOS, RT Linux, or VxWorks)
- Comfortable with Linux OS fundamentals and kernel basics
- Familiarity with actuator control: DC motors, servos, stepper motors, BLDCs with ESCs
- Understanding of BLDC motor control basics: FOC, PWM, current/position control
- Experience with CAN bus, ROS/ROS2, or industrial communication protocols
- Hands-on experience deploying robots or manipulators in the real world
- Experience deploying controllers on embedded hardware (Jetson, STM32, Raspberry Pi)
Why Join Us
We make robot deployment work like SaaS. Currently deploying teleoperation systems for KUKA robots across the world's largest appliance manufacturer. Two founders building the infrastructure layer for industrial robotics. This is the critical path role—we need you before we can collect good training data.